一种大视差下循环指导局部配准的图像拼接方法  

Rolling Guidance Local Alignment for Image Stitching Under Large Parallax

作　　者：汪洋鑫 李凯 沈礼权[1] 安平[1] 邹雪妹[1] 

机构地区：[1]上海先进通信与数据科学研究院,上海大学通信与信息工程学院,上海200444

出　　处：《工业控制计算机》2018年第8期115-117,共3页Industrial Control Computer

摘　　要：在大视差下拼接图像极具挑战。一种以缝隙线质量为导向的局部配准的迭代方法,可以同时快速找到最佳局部区域和缝隙。算法以缝隙的质量为指导迭代地选择局部匹配点。在每次迭代期间,更新已选匹配点和缝隙。接着,再采用保存内容的变换方法来提升配准质量。最后,选择具有最佳缝隙质量的局部配准作为最终配准结果。实验结果表明了该方法在处理大视差场景下的图像拼接时具有有效性和优越性。This paper presents an iterative method for local alignment guided by the quality of the stitching seam,which can quickly find the best local warp and seam.Specifically,local feature matches are iteratively selected guided by the quality of the stitching seam.During each iteration process,we update the selected feature matches and seam.Then,a content-preserving warping method is adopted to refine the roughly aligned region.Finally,this paper select the local alignment result with the best seam quality as the final result.

关 键 词：图像拼接 局部配准 循环指导 视差 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]